User Files from McHazard
Upload All User Files

#67112465632959311 - WL1, 71 coin/34 heart test

In 03:59.69 (14357 frames), 1150 rerecords
Uploaded 10/31/2020 9:53 AM by McHazard (see all 7)
Changes to the previous WIP:
The big heart in the first half of course 4 appears to be the best option because of how much time I spend close to it anyway. Grabbing it costs only 26 frames, compared to the 64 or so it takes before it becomes collectible. There's another heart further in the tunnel that's also worth looking into.
There's a new method in the course 3 revisit to get a return instruction in memory without having to go way above the top of the screen. By breaking a block while there's still a jet flame, ALL of the flame's data is copied into the AAF0 slot, instead of just what gets left behind after it unloads.
As expected, this new setup saves several seconds on the minimap. The run's not officially sub-4 yet because I'm still relying on the accessible debug mode to set my coin and heart count, but I've been spending the past few days investigating how much I can do to manip coins without losing any more time. My estimated maximum is generally hovering around 63, but I can still lose a few frames and still make the same frame rule. We'll see if that's enough.
As for the 36 coin/1 heart setup, it doesn't look viable. My best attempt at crossing the lava pool with only 1 bird kill loses about 150 frames, and I'll lose another 80 or so not being able to despawn the course 3 skull door.

#66957576242994730 - WL1 GEG testing

In 04:07.15 (14803 frames), 0 rerecords
1 comment, 381 downloads
Uploaded 10/24/2020 10:28 AM by McHazard (see all 7)
We have a working assuming no more surprise inaccuracies setup!
There's a lucky coincidence which might be the only reason this works at all; an instruction I'm forced to use corrupts the stack pointer such that a POP will return A8C5, which is right next to the addresses I need to activate debug mode. Of course, that doesn't change the fact that I need to somehow fix the stack pointer before I hit the return instruction, or else everything breaks.
The current setup requires 78 coins and 22 hearts. There are enough blocks in course 4 that I can potentially get all the way there just from random drops, although that won't be fun to manip. As for the hearts, I can pick up a large one during the course 3 revisit without going too far out of my way. I'm still a few short but there's a couple of places I have in mind to pick the rest up. For now I've just used the other debug mode for testing.
Something I didn't see coming was that I now need the waves on the Rice Beach map to be at a specific point in their cycle, and waiting for both that and the 256 frame global timer to line up together costs a ton of time. I thought I could cut out some of that by returning to the world map and re-entering, but that sets the current course ID to hex 30, which is outside the normal range. Trying to enter that course or select a different one will either force the game into an infinite loop or halt it on an illegal instruction.
Other possibilities I'm looking into are 71 coins/34 hearts (requires at least one more big heart) and 36/01 (avoiding that many kills causes a lot of complications), either of which would reduce the wait time on the map screen.

#63426004506127314 - TMNT 4, hopefully for real this time

In 02:15.86 (8165 frames), 0 rerecords
1 comment, 413 downloads
Uploaded 5/18/2020 9:23 AM by McHazard (see all 7)
Yeah, I'm working on this again.
This is 7 frames faster than the WIP from 3 years ago. There are a couple changes but they're pretty small, and mostly serve to get Raph and the enemies into better positions.
My comment on the last WIP said there were variations that were very close together, but I've long forgotten what they actually were. There's a real possibility that I end up optimizing the current strats as much as I can and just move on. Hopefully that doesn't happen, but on the plus side it'd mean getting the TAS finished a year or two faster. :P

#38216693731293348 - CV1, Death -30 frames

In 00:10.92 (656 frames), 0 rerecords
Game: Castlevania ( NES, see all files )
Uploaded 4/9/2017 1:58 AM by McHazard (see all 7)
Savestate is altered from frame 33357 of the published TAS. Once again I've edited it for demonstration purposes, changing the subweapon (address 015B from 11 to 13) and giving myself a few more hearts (address 0071 from 2 to 5).
As expected the axes make quick work of Death, saving 30 frames. This isn't fully optimized yet, it's just the first time I was able to get both of the nearby scythes in position to get extra damage. Getting them in the right place cost some time, and I had to pause a couple times to keep Death from flying off to soon, so there's still room for improvement here.

#38057086532184992 - Stage 12, -8 frames

In 00:34.74 (2088 frames), 0 rerecords
Game: Castlevania ( NES, see all files )
Uploaded 4/1/2017 9:27 PM by McHazard (see all 7)
Note: This movie starts from a hacked savestate for demonstration purposes. Specifically, frame 24512 of the published TAS, but with the subweapon changed (address 015B from 11 to 13).
The April 1 TAS I had planned is going to be late. In the meantime, I've found a few frames in Castlevania. Okay yeah it's real, but "axes are faster" is enough of a joke on its own.
Just like crosses, which aren't available in this stage, axes can multi-hit enemies when passing through two of them. They can't be used as effectively because of their weird arc, but they still help a little because Frankenstein has that hunchback. Who also spits fire.
I extend the movie at the end to regain the holy water in stage 13. If the axes stop being useful here--and I'm not convinced they will--this should keep them from costing any of those frames.

#36193945397944320 - TMNT IV, better stage 2 strats

In 02:15.98 (8172 frames), 0 rerecords
Uploaded 1/7/2017 11:41 PM by McHazard (see all 7)
Now with something that isn't a placeholder in the second half of stage 2. Saved another 122 frames up to when the last enemy unloads. (Also adjusted input so that that's where it ends.)
There's still some variations that are very close together, and I'm not done with positional optimizations either. It'll probably take a while to sort everything out and get this stage finished.

#36046098526977841 - Oh right I was supposed to be doing TMNT IV.

In 02:18.46 (8321 frames), 0 rerecords
Uploaded 1/1/2017 7:53 AM by McHazard (see all 7)
So as it turns out, staying away until I had something to show for my absence was a terrible idea and just resulted in me not working on anything. Never doing that again.
Anyway, this is where I left off work on TMNT IV. It's somewhere around 35-40 seconds faster through the first two stages, though because of the lag differences between emulators I was never able to get an exact frame count.
Everything up to the start of the fence section is finalized... probably. I might need to look over it again. After that, it's a quick test run to get an idea of how many special attacks I'll need to use.
I'd say more, but it's been so long since I've worked on this that I need some time to remember everything.